Supplementary Appendix for At the Edge of War: Frontline Ally Support for the U.S. Military

Contents

  1. Overview - 1
  2. Power Analysis - 1
  3. Descriptive Figures - 3
  4. Tables - 5

Overview

These appendices contain supplementary information for the paper Supplementary Appendix for Outside Threats and Public Perceptions of the U.S. Military in Poland. Herein we provide a number of additional resources related to the project. First, we provide basic information about the survey and data collection procedures. Second, we provide some basic descriptive statistics and information to help readers better understand the data and the distribution of key variables and responses. Third, to save space in the primary manuscript we include all of the model tables for the project here. Fourth, we also include a number of additional figures to help communicate the results of our analysis. Finally, we include a number of diagnostic plots generated from the models we run. In general, we focus on a few specific types of plots and, where necessary, on key variables. For example, traceplots for multilevel multinomial logit models can quickly become both numerous and unwieldy in the confines of a PDF or printed document.

Power Analysis

Before analyzing the data we developed a Bayesian power analysis in an effort to evaluate the probability of correctly identifying true effects versus false positives for the experimental treatment effects. We follow Kruschke (2015) in carrying out this test and implement it in the following steps.

First, we wrote a function that would simulate data that look like the expected sample data. In addition to our survey plan, we used data from Michael A. Allen et al. (2020) and Michael A. Allen et al. (2022) to establish a baseline expectation for what the distribution of the variables should look like.

Second, we generate a set of expected coefficient/effect values for all of the variables in our model. Note that for each variable we allow the expected effect to vary, establishing a mean and standard deviation for the expected effect rather than fixing its value. Where our variables overlap with those included in Michael A. Allen et al. (2022), we use their posterior estimates to generate our expected effect sizes and distributions. Where our variables differ (for example, we include variables that capture respondents’ income sources/occupational fields) we set the expected effects to 0 with a standard deviation of 0.5 to reflect our uncertainty in the parameter values. This does not apply to the treatment variables, which we address more fully below. We also include varying intercepts for the 16 Polish provinces to match our plan to model the actual data using varying intercepts. In general, where we expect an effect we set the standard deviation so it is less than half of the mean beta value.

Third, following this procedure we generate 200 hypothetical data sets for a given sample size value.

Fourth, we chose a set of hypothetical sample size values to evaluate the model’s ability to recover the expected parameter space. Specifically, we choose sample sizes of 1600, 2560, 4800, and 12800. Our actual data are close to 2500, but we chose the other values to assess the model’s performance across a wider range of hypothetical circumstances.

In total, we end up with \(I \times K\) datasets and models, where \(I\) is the number of iterations per sample size (i.e. 200) and \(K\) is the number of different sample sizes (i.e. 4. In our case, we generate 800 sample data sets and run the model a total of 800 times. As in our primary manuscript, we model the hypothetical data using a Bayesian multilevel multinomial logistic regression using {brms} (Bürkner 2017, 2018).

For the treatment values, we do not have strong priors as to what constitute accurate effects. Accordingly, we generate parameters with a couple of considerations. First, given that we are estimating a multinomial logistic regression, the plausible parameter space is fairly constrained. Extreme values (e.g. \(|\beta| > 4\)) are unlikely (except in cases where observations appear to be rare). Second, we look at the effect sizes on similar variables in Michael A. Allen et al. (2020) and Michael A. Allen et al. (2022). Third, we generally expect that the different treatment prompts will increase support for a U.S. presence and decrease opposition. However, we also expect they will yield different magnitudes, with the combined treatment mentioning security concerns and economic benefits yielding the largest of the three. We also view this as an opportunity to evaluate our ability to recover effects of different sizes, and so we set the parameter distributions to values that we think fall within the plausible range, but also run the range of “small” to “large” effects.

Given these considerations, the distributions we use in the power analysis are as follows.

\[ \begin{align} \text{Support} & \begin{cases} Treatment_{SecurtyandEconomic} &\sim N(1.0, 0.3) \\ Treatment_{Security} &\sim N(0.5, 0.1) \\ Treatment_{Economic} &\sim N(0.1, 0.02) \end{cases} \\ \text{Oppose} & \begin{cases} Treatment_{SecurtyandEconomic} &\sim N(-0.8, 0.3) \\ Treatment_{Security} &\sim N(-0.5, 0.2) \\ Treatment_{Economic} &\sim N(-0.1, 0.05) \end{cases} \\ \text{Don't know/Decline to answer} & \begin{cases} Treatment_{SecurtyandEconomic} &\sim N(-0.5, 0.25) \\ Treatment_{Security} &\sim N(-0.2, 0.1) \\ Treatment_{Economic} &\sim N(-0.1, 0.08) \end{cases} \end{align} \]

The following figures show the results of our power analysis. The first figure shows the average \(Pr(Direction)\) score for the treatment variables. The \(Pr(Direction)\) statistic tells us the proportion of the posterior distribution that falls above/below 0 on the same side as the median value. If we had a median coefficient estimate where the median \(\beta = 0.5\) and \(Pr(Direction) = 0.97\), this tells us that there is a 97% chance of a positive effect. An average \(Pr(Direction)\) of 0.90, for example, would therefore tell us that, on average, there is a 90% chance of a positive effect.

It is common to see power analyses presented in terms of what proportion of models’ 95% confidence intervals exclude 0 and demonstrate an effect. We adopt this alternative approach because it allows us to more directly incorporate information about the posterior distribution into our assessment than the conventional frequentist approach.

Figure 1: The average probability of an effect across different treatment and response outcomes.

The second figure shows spaghetti plots whereby the posterior distributions for the treatment effects from the 200 different models are overlaid on top of one another. While this figure does not provide us with a specific statistic as in the case of the previous figure, it does give us a visual check on the distribution of the recovered coefficients and the accompanying uncertainty.

Figure 2: Spaghetti plots dispalying the posterior distributions for treatment coefficients and response outcomes.

In general, the models do a fairly good job of recovering the parameter values we set in our simulation. The average \(Pr(Direction)\) score is above 80% for the first and second treatment variables in the Positive and Negative response categories. The mean expected effect in these cases is approximately 1.0, 0.5, and 0.1 f or the Positive response equation and \(-0.8\), \(-0.5\), and \(-0.1\) for the negative response model. For the Don’t know/Decline response equation we set the expected values to \(-0.5\), \(-0.2\), and \(-0.1\), but we set the standard deviation to a higher value given the relatively low incidence of these responses in existing data and the high level of uncertainty accompanying these responses.

These results indicate that for the largest effect sizes we have a fairly strong chance of recovering the parameter of interest. However, for the smaller effect sizes we are looking at only a 70-75% chance of recovering the parameter of interest. Though this figure may seem high, the smallest value that the \(Pr(Direction)\) statistic can take on is (roughly speaking) 0.50 as it is necessarily tied to the median posterior sample value. Since we do not set any of the Don’t know/Decline coefficients to values close to 0, it makes sense that the posterior samples often have a “larger” portion of their distributions falling below 0. Accordingly, we should be cautious in treating small effects as definitive given our relatively small sample size.

However, our expectations regarding the effects of the treatments prove to be quite wrong. As we discuss in the manuscript, and as we show in the tables below, the treatment effects do not generally correlate strongly with the outcome response. Overall, our expectations regarding the effects of the informational prompts were wide of the mark.

Descriptive Figures

This section includes additional descriptive figures not included in the primary manuscript.

Views of Major Powers

Figure 3 shows the distribution of views of U.S. military personnel deployed to Poland in March of 2023 at the time of our survey. This is a different representation of the 2023 data we show in Figure 1 of the primary manuscript.

Figure 3: Views of U.S. military personnel in Poland among Polish adults

Figure 4 shows the distribution of Polish adults’ views of Poland’s relations with Russia in March of 2023 at the time of our survey.Overwhelmingly respondents indicate that relations between Poland and Russia are “Somewhat hostile” or “Very hostile.”

Figure 4: Views of Russia among Polish adults

Distribution of respondents

Figure 5 shows the number of respondents per province In general, most of the group sizes fall between 100 and 200 respondents per province. The lowest number of respondents per group is 52 (Opolskie) and the highest is 339 (Mazowieckie).

Figure 5: The number of survey respondents per province.

Figure 6 shows the number of respondents per district—the lower level administrative unit below the province. Here we can see substantial skew in the number of respondents per unit. 47 districts produce only one respondent. 64 districts only produce 2 respondents. At the other end of the distribution we have a few districts that produce a vastly disproportionate share of our respondents. 51 respondents come from Łódź, 54 from Poznań, and 150 come from Warszawa. Though we run supplemental analyses using districts as a grouping unit, we do not rely on these estimates to discuss variation in attitudes as a function of geography.

Figure 6: The number of survey respondents per district.

Tables

This section contains a number of tables that provide descriptive insights into the data, and information on the models we run for our analysis.

Balance Tables

Table 1 shows the balance of the predictor variables across the four treatment groups in the experiment. Most of the variables in our models are indicator variables, and so the numbers shown in the columns correspond to the number of respondents who chose a particular response for a particular question. For example, the number of people who respond that they identify as either Male or Female.

The value in the parentheses indicates the percentage of responses that fall into each of the four treatment categories. In general, we expect this value to fall close to 25% for each row.

Last, the final column shows the total number of responses for each category/row.

We do not conduct a formal balance test, but this table helps us to ensure that the randomization procedure worked as intended. In general, we see most response-treatment groups falling at around the 25% mark, which is what we should expect if individuals were randomly assigned to one of the four treatment categories. We see more substantial deviations where the total number of observations for a given response is low. For example, with only 40 total respondents indicating that their primary income source was in the agricultural sector, small differences in the number of people who fall into each treatment group have a larger effect on the percentage value.

The final row shows the mean value and standard deviation (in parentheses) for the ideology score, which is the only ordered integer response variable we included in the survey. Since we mean-center this measure, each category should have a mean of approximately 0 and a standard deviation of 0.5.

Table 1: Balance table for predictors used in primary models.
Treatment Group
Predictor Level Control Security Economic Security and Economic All Groups
Gender
Male 273 (25.4%) 270 (25.1%) 267 (24.8%) 265 (24.7%) 1075
Female 284 (24.1%) 302 (25.7%) 288 (24.5%) 303 (25.7%) 1177
None of the above 0 (0%) 1 (50%) 1 (50%) 0 (0%) 2
Minority
No 456 (25.3%) 455 (25.2%) 449 (24.9%) 442 (24.5%) 1802
Yes 84 (22.5%) 94 (25.2%) 92 (24.7%) 103 (27.6%) 373
Decline to answer 17 (21.5%) 24 (30.4%) 15 (19%) 23 (29.1%) 79
Education
Decline to answer 1 (20%) 1 (20%) 1 (20%) 2 (40%) 5
Higher Education (Bachelor/Engineer) 73 (23.2%) 74 (23.5%) 83 (26.3%) 85 (27%) 315
Higher Education (Master’s degree or higher) 148 (24.7%) 161 (26.9%) 148 (24.7%) 142 (23.7%) 599
Primary Education 24 (34.8%) 19 (27.5%) 12 (17.4%) 14 (20.3%) 69
Secondary Education 238 (25%) 240 (25.2%) 240 (25.2%) 234 (24.6%) 952
Vocational School 73 (23.2%) 78 (24.8%) 72 (22.9%) 91 (29%) 314
Age
18 to 24 years 58 (28.6%) 50 (24.6%) 41 (20.2%) 54 (26.6%) 203
25 to 34 years 114 (26.6%) 109 (25.5%) 108 (25.2%) 97 (22.7%) 428
35 to 44 years 112 (23.7%) 136 (28.8%) 110 (23.3%) 115 (24.3%) 473
45 to 54 years 91 (23.6%) 85 (22.1%) 104 (27%) 105 (27.3%) 385
55 to 64 years 115 (23.3%) 133 (27%) 120 (24.3%) 125 (25.4%) 493
Age 65 or older 67 (24.6%) 60 (22.1%) 73 (26.8%) 72 (26.5%) 272
Income
0 – 43 339 109 (25.6%) 98 (23%) 111 (26.1%) 108 (25.4%) 426
43 340 – 57 187 85 (22.3%) 112 (29.4%) 82 (21.5%) 102 (26.8%) 381
57 188 – 74 062 113 (25.1%) 115 (25.6%) 104 (23.1%) 118 (26.2%) 450
74 063 – 93 937 105 (23.4%) 120 (26.7%) 118 (26.3%) 106 (23.6%) 449
93 938 + 114 (25.3%) 110 (24.4%) 120 (26.7%) 106 (23.6%) 450
Decline to answer 31 (31.6%) 18 (18.4%) 21 (21.4%) 28 (28.6%) 98
Income Source
Agriculture 8 (20%) 15 (37.5%) 6 (15%) 11 (27.5%) 40
Full-time or contract work in the government or public sector 57 (28.9%) 49 (24.9%) 46 (23.4%) 45 (22.8%) 197
Full-time or contract work in the private sector 304 (25.2%) 300 (24.8%) 301 (24.9%) 303 (25.1%) 1208
Other sources 56 (24.1%) 65 (28%) 54 (23.3%) 57 (24.6%) 232
Pension or retirement 103 (22.5%) 113 (24.7%) 121 (26.5%) 120 (26.3%) 457
Self-employed (non-agricultural) 29 (24.2%) 31 (25.8%) 28 (23.3%) 32 (26.7%) 120
Ideology
Ideology -0.024 (0.488) 0.02 (0.499) 0.006 (0.503) -0.003 (0.51) 0 (0.5)

Model Tables

This section contains the tables for the models we run in our analysis. All of the models were run using brms package version 2.21.0 [Bürkner (2017); Bürkner (2018); Stan2023].

  1. Table 2 shows the results of a multinomial logit model where we regress the outcome variable on the treatment group variable.
  2. Table 3 shows the results of our primary multinomial multilevel logit model. This model regresses the outcome response onto the treatment variable and several other predictor variables. Varying intercepts by province.
  3. Table 4 shows the results of our a multinomial multilevel logit model that regresses the outcome response onto the treatment variable and several other predictor variables. Varying intercepts by province and district.
  4. Table 5 shows the results of a model where we use the full six category response variable rather than the four category response used in our primary models.
  5. Table 6 shows the results of a model that replicates our primary multilevel model, but allows the effects of the treatment variable to vary across province.
  6. Table 7 changes the basic model specification slightly and uses the treatment group as the grouping term for the varying intercepts. We also include a variable indicating whether the respondent reported having personal contact with a U.S. service member, and we allow this effect to vary across treatment groups.
  7. Table 8 builds upon our primary model in Table 3 by adding a variable indicating whether the respondent reported having personal contact with a U.S. service member, and an interaction term between the contact variable and the treatment. We also include varying intercepts on province.
  8. Table 9 replicates the models from Table 8 but includes varying intercepts on both province and district.
  9. Table 10 shows the results of a multilevel ordered logit model. Here we take the original six category response variable, drop the “Don’t know/Decline” Responses, and treat the remaining responses as ordered from “Strongly Oppose” to “Strongly Support”.
Table 2: Multinomial logistic regressions with treatment effects and outcome response. Models only include the treatment received by the respondent and their response.
Bivariate
Distance: 100k
Distance: 5k
DKDA Oppose Support DKDA Oppose Support
Treatment
Economic 5.851 0.202 0.243 0.467 0.004 0.063
[2.680, 11.158] [−0.301, 0.706] [−0.091, 0.578] [−0.186, 1.150] [−0.373, 0.369] [−0.250, 0.373]
Security 5.300 0.486 −0.035 −0.053 0.356 0.017
[2.102, 10.600] [0.032, 0.951] [−0.361, 0.284] [−0.793, 0.676] [−0.009, 0.713] [−0.299, 0.328]
Security and Economic 5.618 0.269 0.010 −0.163 0.107 0.049
[2.459, 10.901] [−0.212, 0.758] [−0.319, 0.339] [−0.917, 0.576] [−0.263, 0.475] [−0.262, 0.360]
Intercept −6.898 −0.585 1.536 −1.796 0.109 1.078
[−12.193, −3.786] [−0.941, −0.240] [1.313, 1.773] [−2.328, −1.311] [−0.146, 0.377] [0.864, 1.298]
N 2239 2254
N.Groups 0 0
Table 3: Multilevel multinomial logistic regressions with respondents grouped by province. These model the response as a function of the treatment variables and several predictor variables, with varying intercepts by province.
Base Model
Distance: 100k
Distance: 5k
DKDA Oppose Support DKDA Oppose Support
Treatment
Economic 4.792 0.254 0.204 0.795 0.063 0.046
[2.888, 7.351] [−0.256, 0.760] [−0.149, 0.558] [0.088, 1.515] [−0.320, 0.443] [−0.280, 0.369]
Security 3.948 0.584 −0.062 0.140 0.433 0.019
[2.017, 6.494] [0.110, 1.053] [−0.388, 0.277] [−0.634, 0.915] [0.070, 0.801] [−0.304, 0.343]
Security and Economic 4.359 0.325 0.002 −0.080 0.170 0.085
[2.457, 6.914] [−0.162, 0.815] [−0.337, 0.338] [−0.894, 0.723] [−0.207, 0.547] [−0.240, 0.410]
Age
25-34 −1.799 −0.077 −0.113 −1.394 −0.312 −0.058
[−2.346, −1.248] [−0.404, 0.256] [−0.375, 0.148] [−1.928, −0.854] [−0.605, −0.020] [−0.315, 0.211]
35-44 −1.988 −0.310 0.039 −1.496 −0.617 0.045
[−2.555, −1.435] [−0.641, 0.019] [−0.225, 0.304] [−2.051, −0.949] [−0.906, −0.330] [−0.217, 0.310]
45-54 −1.864 −0.348 0.084 −1.464 −0.694 0.151
[−2.473, −1.268] [−0.699, 0.004] [−0.198, 0.364] [−2.047, −0.878] [−1.006, −0.384] [−0.127, 0.428]
55-64 −1.785 −0.441 0.462 −1.549 −0.925 0.413
[−2.480, −1.096] [−0.846, −0.040] [0.150, 0.771] [−2.205, −0.895] [−1.266, −0.583] [0.115, 0.716]
65+ −2.525 0.194 0.655 −2.230 −0.496 0.564
[−3.693, −1.367] [−0.568, 0.943] [0.135, 1.171] [−3.291, −1.154] [−1.093, 0.092] [0.062, 1.070]
Income
Second Quantile −0.905 0.022 −0.103 −0.914 −0.038 −0.106
[−1.434, −0.374] [−0.302, 0.342] [−0.355, 0.146] [−1.419, −0.410] [−0.319, 0.244] [−0.353, 0.143]
Third Quantile −0.562 0.122 −0.060 −0.709 0.076 −0.030
[−1.074, −0.047] [−0.195, 0.440] [−0.309, 0.194] [−1.206, −0.210] [−0.200, 0.353] [−0.271, 0.212]
Fourth Quantile −0.910 0.012 0.050 −0.949 −0.024 −0.041
[−1.494, −0.328] [−0.334, 0.354] [−0.211, 0.312] [−1.505, −0.402] [−0.317, 0.274] [−0.299, 0.218]
Fifth Quantile −0.757 0.111 0.178 −0.778 0.035 0.137
[−1.434, −0.084] [−0.276, 0.506] [−0.116, 0.475] [−1.429, −0.133] [−0.294, 0.358] [−0.148, 0.422]
Income Decline −0.317 0.636 −0.124 0.302 0.238 −0.318
[−1.583, 0.892] [−0.189, 1.455] [−0.755, 0.539] [−0.660, 1.245] [−0.410, 0.893] [−0.934, 0.303]
Income Source
Public sector contract work 1.335 0.137 0.643 −0.237 0.094 0.870
[−1.122, 4.249] [−1.099, 1.422] [−0.311, 1.546] [−1.944, 1.612] [−0.885, 1.074] [−0.057, 1.780]
Private sector contract work 1.406 0.105 0.629 −0.214 0.261 1.020
[−0.822, 4.256] [−1.005, 1.258] [−0.234, 1.434] [−1.693, 1.468] [−0.631, 1.146] [0.154, 1.864]
Pension or Retirement 2.590 −0.179 0.573 1.198 0.209 1.134
[0.178, 5.569] [−1.449, 1.148] [−0.377, 1.470] [−0.425, 2.980] [−0.805, 1.228] [0.196, 2.064]
Self-employed (non-agricultural) −0.049 −0.320 0.390 −1.259 0.213 1.104
[−3.478, 3.383] [−1.648, 1.035] [−0.583, 1.332] [−4.175, 1.228] [−0.832, 1.274] [0.121, 2.074]
Other sources 1.887 0.090 0.574 −0.151 −0.153 0.429
[−0.377, 4.761] [−1.083, 1.308] [−0.337, 1.441] [−1.678, 1.553] [−1.092, 0.774] [−0.494, 1.324]
Education
Bachelor's degree or Engineer 0.767 0.474 1.933 3.473 0.157 5.177
[−2.595, 4.253] [−2.510, 3.915] [−0.983, 5.347] [−1.098, 11.490] [−2.067, 2.515] [−0.157, 15.042]
Master's degree or higher 0.833 0.253 1.647 3.001 0.371 5.408
[−2.492, 4.354] [−2.715, 3.672] [−1.290, 5.022] [−1.593, 11.073] [−1.839, 2.710] [0.068, 15.253]
Primary Education 1.975 0.078 2.134 5.457 1.176 6.356
[−1.477, 5.517] [−3.134, 3.639] [−0.888, 5.591] [0.876, 13.530] [−1.198, 3.678] [0.920, 16.198]
Secondary Education 0.345 0.305 1.588 3.281 0.458 5.311
[−2.926, 3.776] [−2.650, 3.697] [−1.339, 4.941] [−1.201, 11.283] [−1.737, 2.782] [−0.015, 15.176]
Vocational School 0.921 0.284 1.335 3.684 0.365 5.071
[−2.358, 4.357] [−2.686, 3.706] [−1.601, 4.710] [−0.791, 11.703] [−1.837, 2.701] [−0.233, 14.901]
Ideology
Ideology −0.298 −0.372 0.597 −0.320 −0.302 0.567
[−0.541, −0.057] [−0.498, −0.246] [0.491, 0.703] [−0.566, −0.074] [−0.421, −0.184] [0.460, 0.674]
Minority
Minority: Yes 0.188 0.058 −0.213 −0.039 −0.070 −0.214
[−0.214, 0.595] [−0.157, 0.273] [−0.380, −0.043] [−0.436, 0.351] [−0.262, 0.123] [−0.381, −0.045]
Minority: Decline 2.044 −0.221 −0.450 1.106 −0.583 −0.594
[1.311, 2.765] [−0.877, 0.424] [−0.947, 0.052] [0.396, 1.794] [−1.150, −0.024] [−1.095, −0.093]
Gender
Female 0.605 −0.111 −0.402 0.660 0.013 −0.394
[0.361, 0.849] [−0.243, 0.021] [−0.506, −0.298] [0.423, 0.897] [−0.105, 0.131] [−0.496, −0.294]
None of the Above −12.237 −0.774 −23.846 −12.495 −0.590 −23.047
[−37.690, 1.661] [−4.638, 3.041] [−48.109, −3.720] [−37.985, 1.731] [−4.563, 3.202] [−47.732, −2.780]
Intercept
Intercept −6.835 −0.850 −0.530 −4.417 −0.009 −5.081
[−11.745, −2.335] [−4.447, 2.349] [−3.986, 2.540] [−12.570, 0.413] [−2.508, 2.368] [−14.905, 0.304]
N 2239 2254
N.Groups 16 16
Groups province province
Table 4: Multilevel multinomial logistic regressions with respondents grouped by province and district. These model the response as a function of the treatment variables and several predictor variables, with varying intercepts by province and by district.
District Model
Distance: 100k
Distance: 5k
DKDA Oppose Support DKDA Oppose Support
Treatment
Economic 5.037 0.255 0.212 0.905 0.047 0.038
[3.065, 7.616] [−0.268, 0.774] [−0.144, 0.565] [0.152, 1.681] [−0.336, 0.430] [−0.287, 0.369]
Security 4.039 0.587 −0.065 0.153 0.429 0.012
[2.081, 6.574] [0.100, 1.063] [−0.405, 0.277] [−0.660, 0.965] [0.060, 0.802] [−0.313, 0.337]
Security and Economic 4.569 0.332 0.006 −0.023 0.186 0.092
[2.630, 7.109] [−0.159, 0.831] [−0.338, 0.346] [−0.846, 0.804] [−0.190, 0.565] [−0.230, 0.421]
Age
25-34 −1.931 −0.066 −0.104 −1.497 −0.298 −0.053
[−2.518, −1.351] [−0.391, 0.258] [−0.367, 0.159] [−2.058, −0.935] [−0.599, 0.002] [−0.320, 0.213]
35-44 −2.134 −0.299 0.048 −1.625 −0.602 0.053
[−2.730, −1.540] [−0.624, 0.033] [−0.217, 0.312] [−2.201, −1.055] [−0.901, −0.311] [−0.215, 0.319]
45-54 −2.003 −0.340 0.096 −1.573 −0.683 0.158
[−2.644, −1.379] [−0.690, 0.008] [−0.186, 0.376] [−2.187, −0.970] [−0.999, −0.369] [−0.123, 0.441]
55-64 −1.948 −0.426 0.472 −1.675 −0.909 0.420
[−2.692, −1.220] [−0.824, −0.016] [0.166, 0.785] [−2.368, −0.989] [−1.257, −0.560] [0.116, 0.726]
65+ −2.744 0.217 0.668 −2.361 −0.481 0.571
[−3.979, −1.498] [−0.540, 0.977] [0.137, 1.194] [−3.493, −1.240] [−1.087, 0.128] [0.062, 1.074]
Income
Second Quantile −0.877 0.021 −0.096 −0.928 −0.028 −0.101
[−1.412, −0.330] [−0.307, 0.352] [−0.346, 0.156] [−1.445, −0.408] [−0.308, 0.255] [−0.351, 0.149]
Third Quantile −0.530 0.121 −0.057 −0.712 0.087 −0.024
[−1.065, 0.000] [−0.194, 0.444] [−0.307, 0.192] [−1.216, −0.210] [−0.193, 0.366] [−0.267, 0.219]
Fourth Quantile −0.848 0.008 0.057 −0.927 −0.017 −0.034
[−1.436, −0.256] [−0.340, 0.360] [−0.211, 0.327] [−1.503, −0.357] [−0.313, 0.280] [−0.294, 0.224]
Fifth Quantile −0.688 0.109 0.186 −0.753 0.036 0.142
[−1.381, 0.004] [−0.279, 0.499] [−0.110, 0.480] [−1.419, −0.072] [−0.288, 0.361] [−0.144, 0.426]
Income Decline −0.134 0.647 −0.107 0.391 0.260 −0.307
[−1.456, 1.150] [−0.183, 1.491] [−0.748, 0.569] [−0.629, 1.387] [−0.405, 0.928] [−0.925, 0.326]
Income Source
Public sector contract work 1.285 0.127 0.645 −0.395 0.105 0.855
[−1.308, 4.365] [−1.118, 1.427] [−0.308, 1.562] [−2.208, 1.562] [−0.915, 1.114] [−0.079, 1.776]
Private sector contract work 1.205 0.101 0.633 −0.451 0.272 1.005
[−1.199, 4.170] [−1.011, 1.271] [−0.249, 1.446] [−2.034, 1.330] [−0.657, 1.186] [0.136, 1.867]
Pension or Retirement 2.484 −0.200 0.578 1.014 0.214 1.125
[−0.065, 5.559] [−1.499, 1.126] [−0.389, 1.507] [−0.702, 2.902] [−0.830, 1.243] [0.196, 2.068]
Self-employed (non-agricultural) −0.359 −0.350 0.376 −1.442 0.239 1.099
[−3.890, 3.151] [−1.665, 1.004] [−0.614, 1.326] [−4.449, 1.091] [−0.856, 1.322] [0.103, 2.100]
Other sources 1.715 0.082 0.580 −0.349 −0.138 0.424
[−0.712, 4.694] [−1.105, 1.305] [−0.358, 1.461] [−1.953, 1.459] [−1.123, 0.841] [−0.493, 1.334]
Education
Bachelor's degree or Engineer 0.571 0.485 1.928 3.797 0.063 5.016
[−3.097, 4.342] [−2.531, 3.976] [−0.951, 5.229] [−1.151, 12.212] [−2.202, 2.424] [−0.182, 14.701]
Master's degree or higher 0.551 0.261 1.630 3.286 0.282 5.247
[−3.126, 4.323] [−2.726, 3.758] [−1.226, 4.864] [−1.688, 11.707] [−1.969, 2.642] [0.047, 14.902]
Primary Education 2.030 0.064 2.154 5.849 1.099 6.217
[−1.758, 5.939] [−3.137, 3.734] [−0.764, 5.517] [0.828, 14.299] [−1.309, 3.622] [0.895, 15.911]
Secondary Education 0.039 0.311 1.577 3.541 0.366 5.153
[−3.546, 3.762] [−2.671, 3.793] [−1.284, 4.826] [−1.400, 11.966] [−1.862, 2.719] [−0.019, 14.791]
Vocational School 0.744 0.293 1.336 4.052 0.252 4.908
[−2.836, 4.423] [−2.708, 3.747] [−1.538, 4.618] [−0.831, 12.499] [−2.001, 2.615] [−0.298, 14.578]
Ideology
Ideology −0.314 −0.373 0.600 −0.328 −0.303 0.570
[−0.565, −0.063] [−0.499, −0.249] [0.495, 0.706] [−0.572, −0.084] [−0.419, −0.185] [0.466, 0.674]
Minority
Minority: Yes 0.188 0.059 −0.214 −0.037 −0.068 −0.215
[−0.220, 0.595] [−0.157, 0.272] [−0.384, −0.044] [−0.439, 0.359] [−0.264, 0.127] [−0.382, −0.048]
Minority: Decline 2.292 −0.234 −0.479 1.306 −0.618 −0.610
[1.499, 3.103] [−0.909, 0.414] [−0.980, 0.030] [0.538, 2.066] [−1.190, −0.052] [−1.110, −0.109]
Gender
Female 0.597 −0.110 −0.400 0.666 0.011 −0.394
[0.348, 0.842] [−0.244, 0.022] [−0.504, −0.295] [0.429, 0.905] [−0.107, 0.128] [−0.495, −0.294]
None of the Above −12.133 −0.821 −23.909 −12.448 −0.534 −22.982
[−38.650, 2.220] [−4.802, 3.075] [−48.643, −3.721] [−38.251, 1.763] [−4.339, 3.167] [−48.142, −2.411]
Intercept
Intercept −7.043 −0.895 −0.532 −4.863 0.026 −4.910
[−12.244, −2.297] [−4.560, 2.318] [−3.871, 2.418] [−13.382, 0.380] [−2.504, 2.432] [−14.558, 0.375]
N 2239 2254
N.Groups 16 16
Groups province, province:district province, province:district
Table 5: Multilevel multinomial logistic regressions with respondents grouped by province. These model the response as a function of the treatment variables and several predictor variables, with varying intercepts by province. Here we use the original six response categories rather than the four aggregated categories from the main model.
Full Response Variable
Distance: 100k
Distance: 5k
Stronglysupport Somewhatsupport Somewhatoppose Stronglyoppose DKDA Stronglysupport Somewhatsupport Somewhatoppose Stronglyoppose DKDA
Treatment
Economic 0.162 0.239 −0.013 0.439 5.533 −0.004 0.049 0.115 −0.078 0.561
[−0.228, 0.552] [−0.139, 0.620] [−0.668, 0.637] [−0.240, 1.118] [2.733, 9.945] [−0.385, 0.372] [−0.308, 0.405] [−0.340, 0.568] [−0.558, 0.393] [−0.131, 1.266]
Security 0.094 −0.186 0.369 0.753 4.796 0.138 −0.073 0.426 0.406 −0.064
[−0.284, 0.472] [−0.560, 0.193] [−0.216, 0.963] [0.123, 1.395] [1.971, 9.200] [−0.233, 0.509] [−0.443, 0.298] [−0.012, 0.870] [−0.039, 0.860] [−0.833, 0.707]
Security and Economic 0.113 −0.075 0.047 0.560 5.253 0.225 −0.047 0.192 0.095 −0.172
[−0.257, 0.481] [−0.441, 0.294] [−0.567, 0.653] [−0.078, 1.228] [2.454, 9.665] [−0.141, 0.586] [−0.407, 0.312] [−0.254, 0.641] [−0.373, 0.561] [−0.961, 0.610]
Age
25-34 0.174 0.038 0.081 0.990 0.086 0.326 0.190 −0.071 0.126 1.011
[−0.339, 0.691] [−0.430, 0.513] [−0.661, 0.836] [0.072, 1.995] [−1.011, 1.240] [−0.299, 0.954] [−0.317, 0.697] [−0.605, 0.463] [−0.445, 0.706] [−0.209, 2.390]
35-44 0.649 0.167 0.019 0.828 0.251 0.877 −0.090 −0.432 −0.393 1.346
[0.130, 1.170] [−0.317, 0.648] [−0.754, 0.806] [−0.100, 1.846] [−0.861, 1.417] [0.302, 1.471] [−0.596, 0.417] [−0.973, 0.108] [−0.986, 0.210] [0.146, 2.688]
45-54 0.829 0.335 0.274 1.451 0.667 1.362 0.579 −0.105 0.389 1.760
[0.278, 1.391] [−0.192, 0.865] [−0.559, 1.133] [0.500, 2.489] [−0.503, 1.873] [0.747, 1.996] [0.024, 1.127] [−0.717, 0.503] [−0.251, 1.024] [0.461, 3.198]
55-64 1.280 0.516 −0.266 0.928 1.354 1.414 0.369 −1.212 −0.233 1.751
[0.715, 1.840] [−0.029, 1.059] [−1.207, 0.675] [−0.101, 2.028] [0.211, 2.551] [0.812, 2.035] [−0.172, 0.907] [−1.896, −0.552] [−0.895, 0.428] [0.488, 3.162]
65+ 1.669 1.011 1.304 1.548 0.188 1.637 0.936 −0.241 0.153 0.711
[0.866, 2.481] [0.197, 1.831] [0.042, 2.576] [0.162, 2.964] [−1.556, 1.883] [0.842, 2.451] [0.184, 1.695] [−1.185, 0.688] [−0.824, 1.123] [−0.988, 2.482]
Income
Second Quantile −0.255 0.149 0.373 0.121 −0.056 −0.106 0.179 0.291 0.109 −0.300
[−0.694, 0.190] [−0.282, 0.583] [−0.366, 1.124] [−0.555, 0.803] [−0.933, 0.817] [−0.555, 0.347] [−0.260, 0.620] [−0.230, 0.815] [−0.425, 0.646] [−1.098, 0.482]
Third Quantile −0.209 0.013 0.508 −0.265 −0.207 −0.029 0.149 0.405 −0.134 −0.884
[−0.637, 0.221] [−0.414, 0.440] [−0.196, 1.222] [−0.981, 0.452] [−1.088, 0.659] [−0.459, 0.401] [−0.274, 0.579] [−0.089, 0.901] [−0.682, 0.399] [−1.789, −0.046]
Fourth Quantile 0.354 0.350 0.468 0.066 0.041 0.095 0.216 0.170 0.078 −0.597
[−0.100, 0.808] [−0.101, 0.808] [−0.303, 1.256] [−0.667, 0.795] [−0.907, 0.972] [−0.346, 0.537] [−0.221, 0.645] [−0.358, 0.695] [−0.448, 0.604] [−1.461, 0.241]
Fifth Quantile 0.461 0.432 0.445 0.117 −0.094 0.244 0.318 0.180 −0.151 −0.805
[−0.015, 0.951] [−0.045, 0.921] [−0.391, 1.271] [−0.661, 0.891] [−1.151, 0.924] [−0.210, 0.696] [−0.135, 0.763] [−0.380, 0.734] [−0.733, 0.426] [−1.819, 0.150]
Income Decline −0.342 0.050 0.968 0.267 −0.171 −0.549 −0.080 0.250 0.316 0.481
[−1.066, 0.391] [−0.636, 0.765] [−0.067, 1.987] [−0.822, 1.328] [−1.480, 1.067] [−1.327, 0.212] [−0.792, 0.622] [−0.567, 1.037] [−0.482, 1.104] [−0.489, 1.438]
Income Source
Public sector contract work 0.801 0.443 0.236 0.307 1.282 0.536 1.582 −0.137 0.629 −0.401
[−0.243, 1.833] [−0.615, 1.501] [−1.308, 1.970] [−1.333, 2.165] [−1.181, 4.623] [−0.473, 1.560] [0.250, 3.133] [−1.232, 0.993] [−0.684, 2.115] [−2.073, 1.419]
Private sector contract work 0.628 0.561 0.058 0.452 1.304 0.561 1.825 0.010 0.855 −0.409
[−0.333, 1.574] [−0.394, 1.526] [−1.312, 1.675] [−0.996, 2.180] [−0.928, 4.554] [−0.371, 1.529] [0.556, 3.345] [−0.979, 1.044] [−0.351, 2.238] [−1.819, 1.235]
Pension or Retirement 0.686 0.312 −0.360 0.184 1.851 0.873 1.800 0.137 0.675 0.576
[−0.350, 1.720] [−0.751, 1.368] [−2.019, 1.452] [−1.451, 2.073] [−0.510, 5.175] [−0.145, 1.916] [0.448, 3.369] [−1.014, 1.338] [−0.699, 2.196] [−0.971, 2.342]
Self-employed (non-agricultural) 0.386 0.216 −0.399 −0.051 −0.251 0.628 1.882 0.089 0.664 −1.585
[−0.689, 1.469] [−0.862, 1.311] [−2.118, 1.435] [−1.783, 1.866] [−3.937, 3.532] [−0.461, 1.725] [0.488, 3.503] [−1.112, 1.306] [−0.759, 2.227] [−4.806, 0.982]
Other sources 0.591 0.528 0.270 0.154 1.775 0.176 1.085 −0.359 0.443 −0.148
[−0.417, 1.604] [−0.482, 1.540] [−1.204, 1.941] [−1.435, 1.988] [−0.513, 5.085] [−0.824, 1.198] [−0.245, 2.645] [−1.407, 0.727] [−0.839, 1.874] [−1.615, 1.524]
Education
Bachelor's degree or Engineer 34.514 1.485 −0.178 32.862 −0.194 33.575 33.906 −1.225 34.458 32.448
[2.163, 96.421] [−1.326, 4.981] [−3.134, 3.373] [0.765, 94.330] [−3.373, 3.556] [1.220, 93.729] [1.643, 93.683] [−3.458, 1.108] [1.632, 95.843] [0.345, 92.812]
Master's degree or higher 34.169 1.218 −0.025 32.308 −0.229 33.773 34.159 −0.629 34.358 32.143
[1.735, 96.102] [−1.587, 4.745] [−2.956, 3.483] [0.174, 93.758] [−3.407, 3.499] [1.473, 93.976] [1.935, 94.030] [−2.842, 1.707] [1.527, 95.746] [0.000, 92.274]
Primary Education 34.796 1.473 −1.747 32.717 1.163 34.911 34.850 0.294 34.691 34.729
[2.440, 96.774] [−1.421, 5.048] [−5.858, 2.372] [0.588, 93.926] [−2.129, 4.952] [2.592, 95.253] [2.525, 94.744] [−2.055, 2.776] [1.825, 96.038] [2.613, 95.152]
Secondary Education 34.046 1.233 −0.254 32.655 −0.674 33.645 34.090 −0.538 34.428 32.265
[1.637, 95.907] [−1.556, 4.725] [−3.178, 3.244] [0.578, 94.085] [−3.789, 3.017] [1.349, 93.862] [1.848, 93.928] [−2.730, 1.773] [1.607, 95.762] [0.143, 92.590]
Vocational School 33.965 0.797 −0.190 32.453 −0.014 33.381 33.864 −0.606 34.233 32.693
[1.583, 95.901] [−1.999, 4.279] [−3.103, 3.350] [0.392, 93.930] [−3.118, 3.667] [1.034, 93.593] [1.629, 93.740] [−2.801, 1.721] [1.404, 95.689] [0.597, 93.101]
Ideology
Ideology 0.463 0.154 0.053 0.011 0.208 0.502 0.139 −0.092 0.125 0.091
[0.186, 0.742] [−0.128, 0.438] [−0.400, 0.512] [−0.447, 0.464] [−0.368, 0.789] [0.234, 0.773] [−0.130, 0.409] [−0.424, 0.246] [−0.217, 0.474] [−0.433, 0.627]
Minority
Minority: Yes −0.179 −0.096 0.011 −0.314 0.750 −0.125 −0.194 −0.312 −0.226 0.185
[−0.539, 0.184] [−0.454, 0.262] [−0.563, 0.571] [−0.940, 0.290] [0.049, 1.424] [−0.491, 0.240] [−0.546, 0.163] [−0.737, 0.103] [−0.677, 0.215] [−0.504, 0.850]
Minority: Decline −0.681 −0.547 −0.131 −0.543 1.660 −1.024 −1.059 −1.250 −0.632 0.152
[−1.436, 0.076] [−1.261, 0.173] [−1.208, 0.850] [−1.822, 0.594] [0.680, 2.633] [−1.825, −0.267] [−1.840, −0.313] [−2.186, −0.405] [−1.454, 0.139] [−0.849, 1.077]
Gender
Female −1.166 −0.466 0.055 −0.642 0.549 −1.029 −0.561 0.236 0.054 0.678
[−1.458, −0.880] [−0.759, −0.174] [−0.425, 0.545] [−1.096, −0.187] [−0.101, 1.233] [−1.304, −0.755] [−0.833, −0.288] [−0.109, 0.582] [−0.292, 0.402] [0.087, 1.301]
None of the Above −53.790 −54.520 −52.756 1.117 −50.275 −52.378 −53.059 −52.376 1.189 −51.244
[−150.865, −2.564] [−151.062, −2.971] [−147.885, −1.620] [−2.816, 5.118] [−145.183, 0.752] [−149.992, −1.122] [−150.906, −1.201] [−148.738, −1.331] [−2.591, 5.066] [−146.952, 0.105]
Intercept
Intercept −33.975 −0.990 −1.635 −35.000 −8.978 −34.363 −35.398 0.090 −35.723 −35.723
[−95.861, −1.541] [−4.641, 2.016] [−5.478, 1.709] [−96.394, −2.719] [−15.214, −3.894] [−94.537, −1.962] [−95.304, −3.039] [−2.501, 2.559] [−97.077, −2.859] [−96.301, −3.454]
N 2239 2254
N.Groups 16 16
Groups province province
Table 6: Multilevel multinomial logistic regressions with respondents grouped by province. These model the response as a function of the treatment variables and several predictor variables, with varying intercepts by province. We also allow the effect of the treatment variables to vary by province.
Group Effects
Distance: 100k
Distance: 5k
DKDA Oppose Support DKDA Oppose Support
Treatment
Economic 4.736 0.214 0.199 0.725 0.077 0.024
[2.740, 7.307] [−0.422, 0.828] [−0.215, 0.617] [−0.130, 1.523] [−0.326, 0.494] [−0.334, 0.379]
Security 3.792 0.579 −0.064 −0.034 0.468 0.024
[1.721, 6.404] [−0.067, 1.215] [−0.417, 0.294] [−1.091, 0.866] [0.027, 0.924] [−0.319, 0.362]
Security and Economic 4.306 0.299 0.010 −0.245 0.152 0.092
[2.347, 6.862] [−0.229, 0.821] [−0.341, 0.361] [−1.303, 0.671] [−0.259, 0.554] [−0.242, 0.429]
Age
25-34 −1.849 −0.068 −0.118 −1.426 −0.310 −0.056
[−2.409, −1.288] [−0.395, 0.263] [−0.377, 0.146] [−1.971, −0.882] [−0.591, −0.021] [−0.320, 0.202]
35-44 −2.032 −0.306 0.038 −1.527 −0.617 0.049
[−2.593, −1.461] [−0.632, 0.019] [−0.224, 0.297] [−2.076, −0.981] [−0.902, −0.325] [−0.217, 0.311]
45-54 −1.892 −0.345 0.082 −1.486 −0.696 0.153
[−2.497, −1.291] [−0.694, 0.009] [−0.198, 0.359] [−2.074, −0.901] [−1.009, −0.392] [−0.125, 0.430]
55-64 −1.821 −0.423 0.458 −1.589 −0.922 0.416
[−2.512, −1.125] [−0.828, −0.020] [0.154, 0.761] [−2.257, −0.932] [−1.267, −0.582] [0.115, 0.714]
65+ −2.566 0.230 0.651 −2.283 −0.500 0.566
[−3.751, −1.370] [−0.522, 0.981] [0.131, 1.167] [−3.351, −1.226] [−1.102, 0.094] [0.065, 1.061]
Income
Second Quantile −0.935 0.014 −0.103 −0.938 −0.037 −0.103
[−1.473, −0.408] [−0.311, 0.338] [−0.357, 0.148] [−1.447, −0.429] [−0.319, 0.245] [−0.350, 0.142]
Third Quantile −0.588 0.119 −0.064 −0.722 0.080 −0.030
[−1.103, −0.074] [−0.200, 0.437] [−0.316, 0.186] [−1.216, −0.225] [−0.198, 0.356] [−0.273, 0.212]
Fourth Quantile −0.939 0.004 0.049 −0.962 −0.022 −0.039
[−1.510, −0.350] [−0.339, 0.348] [−0.217, 0.309] [−1.504, −0.412] [−0.317, 0.270] [−0.294, 0.214]
Fifth Quantile −0.796 0.104 0.174 −0.800 0.040 0.137
[−1.473, −0.120] [−0.281, 0.496] [−0.128, 0.470] [−1.441, −0.154] [−0.286, 0.366] [−0.147, 0.421]
Income Decline −0.399 0.646 −0.143 0.290 0.252 −0.319
[−1.700, 0.808] [−0.191, 1.481] [−0.775, 0.533] [−0.698, 1.238] [−0.410, 0.920] [−0.934, 0.305]
Income Source
Public sector contract work 1.344 0.166 0.659 −0.279 0.111 0.882
[−1.154, 4.367] [−1.071, 1.463] [−0.277, 1.568] [−2.040, 1.608] [−0.885, 1.085] [−0.042, 1.783]
Private sector contract work 1.441 0.162 0.640 −0.268 0.288 1.028
[−0.804, 4.353] [−0.958, 1.339] [−0.222, 1.452] [−1.767, 1.463] [−0.607, 1.184] [0.170, 1.879]
Pension or Retirement 2.604 −0.170 0.594 1.164 0.229 1.151
[0.190, 5.608] [−1.461, 1.154] [−0.344, 1.496] [−0.490, 3.027] [−0.789, 1.236] [0.203, 2.090]
Self-employed (non-agricultural) 0.002 −0.279 0.404 −1.345 0.252 1.117
[−3.411, 3.433] [−1.620, 1.084] [−0.560, 1.332] [−4.400, 1.196] [−0.813, 1.308] [0.136, 2.082]
Other sources 1.904 0.141 0.590 −0.223 −0.133 0.443
[−0.393, 4.785] [−1.063, 1.372] [−0.334, 1.467] [−1.763, 1.527] [−1.085, 0.802] [−0.463, 1.343]
Education
Bachelor's degree or Engineer 0.656 0.410 1.968 3.450 0.083 5.146
[−2.699, 4.096] [−2.541, 3.764] [−0.890, 5.286] [−1.110, 11.262] [−2.129, 2.423] [−0.169, 15.078]
Master's degree or higher 0.693 0.173 1.679 2.980 0.299 5.379
[−2.629, 4.116] [−2.778, 3.517] [−1.168, 4.977] [−1.606, 10.786] [−1.906, 2.608] [0.060, 15.268]
Primary Education 1.869 −0.090 2.168 5.494 1.084 6.319
[−1.602, 5.372] [−3.253, 3.440] [−0.779, 5.545] [0.854, 13.273] [−1.289, 3.553] [0.936, 16.219]
Secondary Education 0.189 0.215 1.624 3.256 0.381 5.284
[−3.068, 3.575] [−2.713, 3.526] [−1.217, 4.929] [−1.206, 11.072] [−1.807, 2.666] [−0.015, 15.166]
Vocational School 0.769 0.212 1.360 3.672 0.294 5.043
[−2.523, 4.137] [−2.738, 3.525] [−1.507, 4.662] [−0.805, 11.486] [−1.893, 2.613] [−0.265, 14.929]
Ideology
Ideology −0.301 −0.375 0.600 −0.324 −0.304 0.569
[−0.543, −0.060] [−0.503, −0.248] [0.494, 0.704] [−0.567, −0.083] [−0.423, −0.185] [0.467, 0.674]
Minority
Minority: Yes 0.188 0.073 −0.217 −0.042 −0.067 −0.213
[−0.216, 0.597] [−0.144, 0.294] [−0.385, −0.048] [−0.434, 0.350] [−0.263, 0.128] [−0.381, −0.048]
Minority: Decline 2.093 −0.213 −0.461 1.137 −0.602 −0.595
[1.340, 2.844] [−0.893, 0.454] [−0.963, 0.054] [0.421, 1.830] [−1.160, −0.053] [−1.101, −0.085]
Gender
Female 0.604 −0.112 −0.402 0.660 0.014 −0.395
[0.361, 0.850] [−0.243, 0.020] [−0.505, −0.300] [0.423, 0.896] [−0.104, 0.133] [−0.496, −0.293]
None of the Above −11.941 −0.906 −23.822 −12.601 −0.628 −23.180
[−37.504, 2.020] [−4.808, 2.977] [−47.890, −3.890] [−39.417, 1.740] [−4.451, 3.056] [−47.735, −2.540]
Intercept
Intercept −6.720 −0.832 −0.570 −4.332 0.027 −5.064
[−11.737, −2.220] [−4.352, 2.332] [−3.945, 2.422] [−12.206, 0.498] [−2.433, 2.372] [−14.963, 0.319]
N 2239 2254
N.Groups 16 16
Groups province province
Table 7: Multilevel multinomial logistic regressions with respondents grouped by treatment groups These model the response as a function of several predictor variables, with varying intercepts by province. We also allow the effect of contact to vary by treatment group.
Personal Contact with Treatment Groupings
Distance: 100k
Distance: 5k
DKDA Oppose Support DKDA Oppose Support
Contact
Personal Contact: Yes −3.302 −0.024 0.701 −3.763 −0.382 0.677
[−11.204, 0.727] [−1.550, 1.445] [−0.290, 1.722] [−12.206, 0.063] [−1.357, 0.589] [0.047, 1.323]
Personal Contact: Don't know/Decline 0.410 −0.882 −0.746 0.463 −1.038 −1.035
[−3.028, 3.288] [−4.668, 2.431] [−2.839, 1.704] [−1.298, 2.279] [−5.261, 2.445] [−2.558, 0.554]
Age
25-34 −1.798 −0.064 −0.091 −1.425 −0.323 −0.027
[−2.359, −1.242] [−0.396, 0.267] [−0.359, 0.178] [−1.965, −0.896] [−0.618, −0.029] [−0.294, 0.239]
35-44 −1.977 −0.294 0.071 −1.538 −0.638 0.097
[−2.533, −1.410] [−0.626, 0.036] [−0.195, 0.339] [−2.080, −0.999] [−0.932, −0.344] [−0.171, 0.365]
45-54 −1.917 −0.336 0.099 −1.523 −0.705 0.189
[−2.517, −1.310] [−0.691, 0.016] [−0.186, 0.377] [−2.105, −0.944] [−1.014, −0.397] [−0.097, 0.471]
55-64 −1.833 −0.432 0.480 −1.597 −0.950 0.458
[−2.535, −1.123] [−0.836, −0.033] [0.171, 0.789] [−2.258, −0.944] [−1.299, −0.603] [0.149, 0.760]
65+ −2.543 0.221 0.669 −2.233 −0.510 0.597
[−3.736, −1.320] [−0.519, 0.962] [0.146, 1.188] [−3.305, −1.169] [−1.123, 0.078] [0.074, 1.098]
Income
Second Quantile −0.887 0.027 −0.116 −0.903 −0.033 −0.104
[−1.418, −0.358] [−0.295, 0.351] [−0.368, 0.133] [−1.406, −0.399] [−0.314, 0.247] [−0.350, 0.141]
Third Quantile −0.511 0.127 −0.054 −0.695 0.071 −0.004
[−1.031, 0.010] [−0.194, 0.453] [−0.303, 0.196] [−1.187, −0.201] [−0.209, 0.342] [−0.248, 0.238]
Fourth Quantile −0.878 0.022 0.051 −0.931 −0.022 −0.026
[−1.457, −0.296] [−0.316, 0.365] [−0.213, 0.319] [−1.495, −0.368] [−0.319, 0.270] [−0.286, 0.229]
Fifth Quantile −0.705 0.132 0.203 −0.761 0.041 0.170
[−1.371, −0.038] [−0.254, 0.521] [−0.096, 0.500] [−1.401, −0.119] [−0.289, 0.371] [−0.115, 0.453]
Income Decline −0.337 0.595 −0.066 0.304 0.225 −0.214
[−1.609, 0.885] [−0.228, 1.409] [−0.713, 0.607] [−0.661, 1.244] [−0.436, 0.897] [−0.840, 0.423]
Income Source
Public sector contract work 1.621 0.131 0.682 −0.277 −0.054 0.877
[−0.842, 4.607] [−1.111, 1.437] [−0.301, 1.617] [−2.042, 1.590] [−1.054, 0.944] [−0.068, 1.819]
Private sector contract work 1.581 0.100 0.642 −0.251 0.152 1.021
[−0.643, 4.422] [−1.019, 1.287] [−0.255, 1.486] [−1.765, 1.455] [−0.764, 1.060] [0.146, 1.887]
Pension or Retirement 2.748 −0.194 0.625 1.089 0.083 1.185
[0.371, 5.727] [−1.462, 1.133] [−0.341, 1.562] [−0.576, 2.895] [−0.956, 1.109] [0.227, 2.136]
Self-employed (non-agricultural) 0.197 −0.269 0.441 −1.292 0.158 1.134
[−3.294, 3.629] [−1.599, 1.088] [−0.559, 1.436] [−4.343, 1.264] [−0.907, 1.232] [0.126, 2.136]
Other sources 1.959 0.122 0.628 −0.286 −0.232 0.465
[−0.291, 4.840] [−1.059, 1.373] [−0.318, 1.554] [−1.850, 1.422] [−1.200, 0.718] [−0.464, 1.389]
Education
Bachelor's degree or Engineer 0.823 0.495 1.816 3.939 0.072 4.956
[−2.529, 4.314] [−2.437, 3.930] [−1.053, 5.171] [−0.680, 11.736] [−2.141, 2.377] [−0.392, 14.850]
Master's degree or higher 0.945 0.309 1.554 3.490 0.297 5.209
[−2.344, 4.416] [−2.630, 3.728] [−1.321, 4.880] [−1.154, 11.337] [−1.896, 2.632] [−0.147, 15.089]
Primary Education 2.089 0.340 2.233 5.828 1.139 6.252
[−1.377, 5.662] [−2.781, 3.956] [−0.744, 5.675] [1.143, 13.742] [−1.231, 3.600] [0.785, 16.163]
Secondary Education 0.552 0.397 1.525 3.815 0.386 5.119
[−2.667, 3.978] [−2.522, 3.771] [−1.343, 4.855] [−0.736, 11.655] [−1.809, 2.676] [−0.226, 14.987]
Vocational School 1.020 0.390 1.258 4.154 0.317 4.874
[−2.245, 4.442] [−2.543, 3.781] [−1.611, 4.571] [−0.367, 12.016] [−1.885, 2.631] [−0.473, 14.772]
Ideology
Ideology −0.295 −0.369 0.594 −0.319 −0.297 0.564
[−0.544, −0.050] [−0.496, −0.243] [0.487, 0.702] [−0.562, −0.075] [−0.416, −0.178] [0.460, 0.667]
Minority
Minority: Yes 0.179 0.050 −0.228 −0.051 −0.053 −0.238
[−0.228, 0.581] [−0.166, 0.266] [−0.396, −0.056] [−0.455, 0.344] [−0.251, 0.143] [−0.406, −0.069]
Minority: Decline 2.028 −0.213 −0.470 1.018 −0.530 −0.618
[1.308, 2.750] [−0.879, 0.427] [−0.966, 0.037] [0.322, 1.715] [−1.102, 0.031] [−1.124, −0.112]
Gender
Female 0.599 −0.102 −0.392 0.660 0.012 −0.380
[0.354, 0.841] [−0.235, 0.030] [−0.496, −0.288] [0.424, 0.896] [−0.102, 0.130] [−0.481, −0.280]
None of the Above −11.892 0.051 −23.998 −12.464 −0.022 −23.226
[−37.919, 1.907] [−3.953, 3.917] [−48.802, −3.576] [−39.008, 1.352] [−3.984, 3.779] [−47.577, −2.376]
Intercept
Intercept −3.215 −0.643 −0.529 −4.451 0.405 −4.984
[−8.444, 1.936] [−4.285, 2.505] [−3.983, 2.456] [−12.461, 0.549] [−2.081, 2.857] [−14.835, 0.445]
N 2239 2254
N.Groups 0 0
Groups treatment_group treatment_group
Table 8: Multilevel multinomial logistic regressions with respondents grouped by province. These model the response as a function of the treatment variables and several predictor variables, with varying intercepts by province. Here we interact the contact variable with the treatment variable to see if the effect of the treatment is conditioned by reported personal contact.
Contact and Treatment Interaction
Distance: 100k
Distance: 5k
DKDA Oppose Support DKDA Oppose Support
Treatment
Economic 5.408 −0.003 0.157 0.855 −0.117 −0.023
[3.011, 8.801] [−0.550, 0.549] [−0.218, 0.531] [0.106, 1.627] [−0.524, 0.293] [−0.360, 0.323]
Security 4.482 0.566 −0.044 0.245 0.486 0.052
[2.085, 7.877] [0.064, 1.068] [−0.410, 0.325] [−0.609, 1.105] [0.093, 0.886] [−0.307, 0.413]
Security and Economic 4.710 0.077 −0.145 −0.167 0.098 0.013
[2.347, 8.127] [−0.445, 0.594] [−0.500, 0.218] [−1.056, 0.714] [−0.296, 0.496] [−0.337, 0.362]
Contact
Personal Contact: Yes −1.471 −0.560 1.123 −0.414 −0.002 0.939
[−14.974, 6.155] [−2.470, 1.099] [0.195, 2.230] [−3.692, 1.841] [−0.988, 1.016] [0.176, 1.787]
Personal Contact: Don't know/Decline −5.892 −9.251 −3.117 −0.992 −39.485 −2.526
[−24.390, 3.132] [−25.754, −1.783] [−5.068, −1.592] [−3.132, 0.829] [−69.607, −10.939] [−4.423, −1.002]
Interactions
Security X Personal Contact 0.120 −0.955 −1.065 −58.163 −1.347 −0.778
[−8.558, 13.954] [−3.261, 1.397] [−2.381, 0.149] [−87.122, −30.444] [−2.805, 0.069] [−1.869, 0.270]
Economic X Personal Contact −17.392 1.122 −0.530 −28.454 0.366 0.058
[−42.759, 3.250] [−1.005, 3.424] [−1.937, 0.884] [−56.755, −4.297] [−1.093, 1.835] [−1.129, 1.282]
Security and Economic X Personal Contact −6.221 1.685 0.354 −13.801 −0.506 −0.114
[−22.339, 9.281] [−0.418, 3.995] [−1.135, 1.877] [−36.455, −0.065] [−1.934, 0.911] [−1.219, 0.997]
Security X Personal Contact: Don't know/Decline 5.238 7.828 2.440 1.901 38.198 2.107
[−4.272, 23.859] [−0.199, 24.364] [0.417, 4.800] [−0.788, 4.725] [9.650, 68.331] [−0.034, 4.488]
Economic X Personal Contact: Don't know/Decline 6.212 11.072 3.120 1.603 41.746 2.242
[−3.986, 25.061] [2.855, 27.718] [0.447, 6.457] [−2.832, 5.779] [12.930, 71.996] [−0.881, 5.756]
Security and Economic X Personal Contact: Don't know/Decline 15.208 15.726 10.296 2.957 40.090 2.313
[1.952, 37.907] [3.238, 37.154] [2.878, 26.031] [0.025, 6.094] [11.412, 70.286] [−0.137, 5.021]
Age
25-34 −1.819 −0.037 −0.052 −1.340 −0.334 −0.016
[−2.388, −1.243] [−0.371, 0.299] [−0.326, 0.221] [−1.910, −0.776] [−0.628, −0.036] [−0.287, 0.260]
35-44 −2.066 −0.299 0.105 −1.124 −0.681 0.110
[−2.662, −1.476] [−0.641, 0.039] [−0.167, 0.380] [−1.734, −0.509] [−0.985, −0.385] [−0.164, 0.387]
45-54 −2.047 −0.344 0.117 −0.926 −0.726 0.211
[−2.690, −1.411] [−0.707, 0.021] [−0.171, 0.409] [−1.610, −0.250] [−1.046, −0.411] [−0.073, 0.494]
55-64 −2.023 −0.501 0.455 −1.441 −1.123 0.426
[−2.746, −1.294] [−0.919, −0.091] [0.139, 0.770] [−2.138, −0.749] [−1.500, −0.747] [0.120, 0.736]
65+ −2.626 0.162 0.669 −2.175 −0.657 0.562
[−3.813, −1.400] [−0.592, 0.923] [0.148, 1.190] [−3.261, −1.091] [−1.274, −0.042] [0.062, 1.062]
Income
Second Quantile −0.983 −0.034 −0.147 −1.393 −0.225 −0.178
[−1.549, −0.425] [−0.366, 0.296] [−0.403, 0.106] [−1.995, −0.797] [−0.531, 0.083] [−0.427, 0.073]
Third Quantile −0.457 0.130 −0.041 −0.835 −0.038 −0.047
[−1.012, 0.098] [−0.192, 0.459] [−0.290, 0.207] [−1.385, −0.291] [−0.320, 0.246] [−0.290, 0.200]
Fourth Quantile −1.067 0.002 0.030 −1.046 −0.162 −0.076
[−1.682, −0.463] [−0.343, 0.347] [−0.235, 0.293] [−1.613, −0.471] [−0.472, 0.149] [−0.335, 0.181]
Fifth Quantile −0.980 0.112 0.161 −1.102 −0.051 0.131
[−1.707, −0.250] [−0.283, 0.508] [−0.137, 0.462] [−1.813, −0.401] [−0.384, 0.285] [−0.160, 0.420]
Income Decline −0.367 0.674 −0.073 0.202 0.271 −0.212
[−1.690, 0.898] [−0.165, 1.513] [−0.728, 0.620] [−0.784, 1.147] [−0.407, 0.957] [−0.855, 0.438]
Income Source
Public sector contract work 1.528 0.037 0.627 −0.316 −0.092 0.821
[−0.995, 4.523] [−1.257, 1.362] [−0.359, 1.568] [−2.072, 1.558] [−1.128, 0.912] [−0.137, 1.763]
Private sector contract work 1.473 0.019 0.592 −0.359 0.108 0.967
[−0.772, 4.341] [−1.139, 1.242] [−0.327, 1.443] [−1.912, 1.353] [−0.838, 1.031] [0.061, 1.857]
Pension or Retirement 2.703 −0.252 0.585 0.957 0.054 1.142
[0.295, 5.665] [−1.589, 1.126] [−0.419, 1.541] [−0.730, 2.811] [−0.996, 1.096] [0.169, 2.107]
Self-employed (non-agricultural) 0.068 −0.304 0.420 −1.302 0.162 1.118
[−3.497, 3.537] [−1.665, 1.075] [−0.607, 1.406] [−4.271, 1.192] [−0.938, 1.273] [0.092, 2.135]
Other sources 1.943 0.061 0.591 −0.359 −0.266 0.415
[−0.345, 4.822] [−1.173, 1.326] [−0.377, 1.518] [−1.933, 1.385] [−1.254, 0.718] [−0.540, 1.346]
Education
Bachelor's degree or Engineer 0.699 0.249 1.727 4.381 −0.057 4.965
[−2.670, 4.242] [−2.787, 3.735] [−1.180, 5.086] [−0.735, 13.390] [−2.322, 2.311] [−0.337, 15.052]
Master's degree or higher 0.877 0.068 1.466 3.945 0.178 5.218
[−2.499, 4.385] [−2.955, 3.532] [−1.426, 4.782] [−1.222, 12.932] [−2.079, 2.541] [−0.102, 15.251]
Primary Education 2.009 0.106 2.187 6.452 1.128 6.348
[−1.451, 5.635] [−3.116, 3.782] [−0.800, 5.576] [1.312, 15.445] [−1.296, 3.676] [0.964, 16.496]
Secondary Education 0.445 0.151 1.451 4.260 0.295 5.151
[−2.860, 3.896] [−2.862, 3.584] [−1.443, 4.784] [−0.804, 13.211] [−1.943, 2.641] [−0.144, 15.205]
Vocational School 0.894 0.139 1.176 4.536 0.191 4.883
[−2.426, 4.334] [−2.833, 3.596] [−1.686, 4.501] [−0.546, 13.453] [−2.069, 2.552] [−0.415, 14.963]
Ideology
Ideology −0.280 −0.373 0.597 0.006 −0.265 0.580
[−0.534, −0.027] [−0.501, −0.244] [0.488, 0.703] [−0.287, 0.301] [−0.386, −0.144] [0.477, 0.685]
Minority
Minority: Yes 0.344 0.038 −0.239 −0.328 −0.050 −0.254
[−0.088, 0.782] [−0.181, 0.255] [−0.411, −0.067] [−0.819, 0.157] [−0.249, 0.149] [−0.422, −0.084]
Minority: Decline 2.065 −0.269 −0.498 0.884 −0.816 −0.735
[1.311, 2.808] [−0.965, 0.407] [−1.014, 0.016] [0.165, 1.590] [−1.429, −0.218] [−1.245, −0.221]
Gender
Female 0.648 −0.065 −0.374 0.794 0.088 −0.367
[0.360, 0.936] [−0.199, 0.070] [−0.481, −0.269] [0.496, 1.098] [−0.040, 0.218] [−0.468, −0.265]
None of the Above −11.613 −0.162 −23.513 −13.216 −0.132 −23.528
[−37.096, 2.346] [−4.241, 3.906] [−48.566, −3.129] [−39.510, 1.305] [−4.065, 3.587] [−48.506, −2.588]
Intercept
Intercept −7.408 −0.474 −0.390 −5.221 0.546 −4.920
[−12.805, −2.546] [−4.107, 2.745] [−3.830, 2.622] [−14.337, 0.147] [−1.962, 2.972] [−14.981, 0.462]
N 2239 2254
N.Groups 16 16
Groups province province
Table 9: Multilevel multinomial logistic regressions with respondents grouped by province and district. These model the response as a function of the treatment variables and several predictor variables, with varying intercepts by province and by district. Here we interact the contact variable with the treatment variable to see if the effect of the treatment is conditioned by reported personal contact.
Contact and Treatment Interaction
Distance: 100k
Distance: 5k
DKDA Oppose Support DKDA Oppose Support
Treatment
Economic 5.408 −0.003 0.157 0.855 −0.117 −0.023
[3.011, 8.801] [−0.550, 0.549] [−0.218, 0.531] [0.106, 1.627] [−0.524, 0.293] [−0.360, 0.323]
Security 4.482 0.566 −0.044 0.245 0.486 0.052
[2.085, 7.877] [0.064, 1.068] [−0.410, 0.325] [−0.609, 1.105] [0.093, 0.886] [−0.307, 0.413]
Security and Economic 4.710 0.077 −0.145 −0.167 0.098 0.013
[2.347, 8.127] [−0.445, 0.594] [−0.500, 0.218] [−1.056, 0.714] [−0.296, 0.496] [−0.337, 0.362]
Contact
Personal Contact: Yes −1.471 −0.560 1.123 −0.414 −0.002 0.939
[−14.974, 6.155] [−2.470, 1.099] [0.195, 2.230] [−3.692, 1.841] [−0.988, 1.016] [0.176, 1.787]
Personal Contact: Don't know/Decline −5.892 −9.251 −3.117 −0.992 −39.485 −2.526
[−24.390, 3.132] [−25.754, −1.783] [−5.068, −1.592] [−3.132, 0.829] [−69.607, −10.939] [−4.423, −1.002]
Interactions
Security X Personal Contact 0.120 −0.955 −1.065 −58.163 −1.347 −0.778
[−8.558, 13.954] [−3.261, 1.397] [−2.381, 0.149] [−87.122, −30.444] [−2.805, 0.069] [−1.869, 0.270]
Economic X Personal Contact −17.392 1.122 −0.530 −28.454 0.366 0.058
[−42.759, 3.250] [−1.005, 3.424] [−1.937, 0.884] [−56.755, −4.297] [−1.093, 1.835] [−1.129, 1.282]
Security and Economic X Personal Contact −6.221 1.685 0.354 −13.801 −0.506 −0.114
[−22.339, 9.281] [−0.418, 3.995] [−1.135, 1.877] [−36.455, −0.065] [−1.934, 0.911] [−1.219, 0.997]
Security X Personal Contact: Don't know/Decline 5.238 7.828 2.440 1.901 38.198 2.107
[−4.272, 23.859] [−0.199, 24.364] [0.417, 4.800] [−0.788, 4.725] [9.650, 68.331] [−0.034, 4.488]
Economic X Personal Contact: Don't know/Decline 6.212 11.072 3.120 1.603 41.746 2.242
[−3.986, 25.061] [2.855, 27.718] [0.447, 6.457] [−2.832, 5.779] [12.930, 71.996] [−0.881, 5.756]
Security and Economic X Personal Contact: Don't know/Decline 15.208 15.726 10.296 2.957 40.090 2.313
[1.952, 37.907] [3.238, 37.154] [2.878, 26.031] [0.025, 6.094] [11.412, 70.286] [−0.137, 5.021]
Age
25-34 −1.819 −0.037 −0.052 −1.340 −0.334 −0.016
[−2.388, −1.243] [−0.371, 0.299] [−0.326, 0.221] [−1.910, −0.776] [−0.628, −0.036] [−0.287, 0.260]
35-44 −2.066 −0.299 0.105 −1.124 −0.681 0.110
[−2.662, −1.476] [−0.641, 0.039] [−0.167, 0.380] [−1.734, −0.509] [−0.985, −0.385] [−0.164, 0.387]
45-54 −2.047 −0.344 0.117 −0.926 −0.726 0.211
[−2.690, −1.411] [−0.707, 0.021] [−0.171, 0.409] [−1.610, −0.250] [−1.046, −0.411] [−0.073, 0.494]
55-64 −2.023 −0.501 0.455 −1.441 −1.123 0.426
[−2.746, −1.294] [−0.919, −0.091] [0.139, 0.770] [−2.138, −0.749] [−1.500, −0.747] [0.120, 0.736]
65+ −2.626 0.162 0.669 −2.175 −0.657 0.562
[−3.813, −1.400] [−0.592, 0.923] [0.148, 1.190] [−3.261, −1.091] [−1.274, −0.042] [0.062, 1.062]
Income
Second Quantile −0.983 −0.034 −0.147 −1.393 −0.225 −0.178
[−1.549, −0.425] [−0.366, 0.296] [−0.403, 0.106] [−1.995, −0.797] [−0.531, 0.083] [−0.427, 0.073]
Third Quantile −0.457 0.130 −0.041 −0.835 −0.038 −0.047
[−1.012, 0.098] [−0.192, 0.459] [−0.290, 0.207] [−1.385, −0.291] [−0.320, 0.246] [−0.290, 0.200]
Fourth Quantile −1.067 0.002 0.030 −1.046 −0.162 −0.076
[−1.682, −0.463] [−0.343, 0.347] [−0.235, 0.293] [−1.613, −0.471] [−0.472, 0.149] [−0.335, 0.181]
Fifth Quantile −0.980 0.112 0.161 −1.102 −0.051 0.131
[−1.707, −0.250] [−0.283, 0.508] [−0.137, 0.462] [−1.813, −0.401] [−0.384, 0.285] [−0.160, 0.420]
Income Decline −0.367 0.674 −0.073 0.202 0.271 −0.212
[−1.690, 0.898] [−0.165, 1.513] [−0.728, 0.620] [−0.784, 1.147] [−0.407, 0.957] [−0.855, 0.438]
Income Source
Public sector contract work 1.528 0.037 0.627 −0.316 −0.092 0.821
[−0.995, 4.523] [−1.257, 1.362] [−0.359, 1.568] [−2.072, 1.558] [−1.128, 0.912] [−0.137, 1.763]
Private sector contract work 1.473 0.019 0.592 −0.359 0.108 0.967
[−0.772, 4.341] [−1.139, 1.242] [−0.327, 1.443] [−1.912, 1.353] [−0.838, 1.031] [0.061, 1.857]
Pension or Retirement 2.703 −0.252 0.585 0.957 0.054 1.142
[0.295, 5.665] [−1.589, 1.126] [−0.419, 1.541] [−0.730, 2.811] [−0.996, 1.096] [0.169, 2.107]
Self-employed (non-agricultural) 0.068 −0.304 0.420 −1.302 0.162 1.118
[−3.497, 3.537] [−1.665, 1.075] [−0.607, 1.406] [−4.271, 1.192] [−0.938, 1.273] [0.092, 2.135]
Other sources 1.943 0.061 0.591 −0.359 −0.266 0.415
[−0.345, 4.822] [−1.173, 1.326] [−0.377, 1.518] [−1.933, 1.385] [−1.254, 0.718] [−0.540, 1.346]
Education
Bachelor's degree or Engineer 0.699 0.249 1.727 4.381 −0.057 4.965
[−2.670, 4.242] [−2.787, 3.735] [−1.180, 5.086] [−0.735, 13.390] [−2.322, 2.311] [−0.337, 15.052]
Master's degree or higher 0.877 0.068 1.466 3.945 0.178 5.218
[−2.499, 4.385] [−2.955, 3.532] [−1.426, 4.782] [−1.222, 12.932] [−2.079, 2.541] [−0.102, 15.251]
Primary Education 2.009 0.106 2.187 6.452 1.128 6.348
[−1.451, 5.635] [−3.116, 3.782] [−0.800, 5.576] [1.312, 15.445] [−1.296, 3.676] [0.964, 16.496]
Secondary Education 0.445 0.151 1.451 4.260 0.295 5.151
[−2.860, 3.896] [−2.862, 3.584] [−1.443, 4.784] [−0.804, 13.211] [−1.943, 2.641] [−0.144, 15.205]
Vocational School 0.894 0.139 1.176 4.536 0.191 4.883
[−2.426, 4.334] [−2.833, 3.596] [−1.686, 4.501] [−0.546, 13.453] [−2.069, 2.552] [−0.415, 14.963]
Ideology
Ideology −0.280 −0.373 0.597 0.006 −0.265 0.580
[−0.534, −0.027] [−0.501, −0.244] [0.488, 0.703] [−0.287, 0.301] [−0.386, −0.144] [0.477, 0.685]
Minority
Minority: Yes 0.344 0.038 −0.239 −0.328 −0.050 −0.254
[−0.088, 0.782] [−0.181, 0.255] [−0.411, −0.067] [−0.819, 0.157] [−0.249, 0.149] [−0.422, −0.084]
Minority: Decline 2.065 −0.269 −0.498 0.884 −0.816 −0.735
[1.311, 2.808] [−0.965, 0.407] [−1.014, 0.016] [0.165, 1.590] [−1.429, −0.218] [−1.245, −0.221]
Gender
Female 0.648 −0.065 −0.374 0.794 0.088 −0.367
[0.360, 0.936] [−0.199, 0.070] [−0.481, −0.269] [0.496, 1.098] [−0.040, 0.218] [−0.468, −0.265]
None of the Above −11.613 −0.162 −23.513 −13.216 −0.132 −23.528
[−37.096, 2.346] [−4.241, 3.906] [−48.566, −3.129] [−39.510, 1.305] [−4.065, 3.587] [−48.506, −2.588]
Intercept
Intercept −7.408 −0.474 −0.390 −5.221 0.546 −4.920
[−12.805, −2.546] [−4.107, 2.745] [−3.830, 2.622] [−14.337, 0.147] [−1.962, 2.972] [−14.981, 0.462]
N 2239 2254
N.Groups 16 16
Groups province province
Table 10: Multilevel ordered logistic regressions with respondents grouped by province These model the response as a function of the treatment variables and several predictor variables, with varying intercepts by province.
 100 km  5 km
Treatment
Economic −0.013 −0.057
[−0.195, 0.168] [−0.233, 0.122]
Security −0.087 −0.218
[−0.266, 0.093] [−0.395, −0.042]
Security and Economic −0.014 0.018
[−0.195, 0.169] [−0.156, 0.195]
Age
25-34 −0.085 0.118
[−0.337, 0.165] [−0.124, 0.363]
35-44 0.297 0.639
[0.043, 0.547] [0.395, 0.884]
45-54 0.232 0.689
[−0.035, 0.498] [0.426, 0.951]
55-64 0.724 1.121
[0.456, 0.994] [0.861, 1.376]
65+ 0.594 0.937
[0.233, 0.957] [0.591, 1.285]
Income
Second Quantile −0.288 −0.184
[−0.508, −0.068] [−0.398, 0.030]
Third Quantile −0.186 −0.097
[−0.400, 0.024] [−0.304, 0.112]
Fourth Quantile 0.156 −0.007
[−0.059, 0.371] [−0.219, 0.204]
Fifth Quantile 0.228 0.180
[0.003, 0.455] [−0.039, 0.399]
Income Decline −0.446 −0.474
[−0.802, −0.094] [−0.816, −0.134]
Income Source
Public sector contract work 0.533 0.398
[−0.016, 1.070] [−0.141, 0.922]
Private sector contract work 0.399 0.320
[−0.111, 0.909] [−0.182, 0.810]
Pension or Retirement 0.595 0.598
[0.052, 1.145] [0.062, 1.126]
Self-employed (non-agricultural) 0.376 0.418
[−0.189, 0.950] [−0.152, 0.972]
Other sources 0.378 0.228
[−0.155, 0.908] [−0.294, 0.747]
Education
Bachelor's degree or Engineer 0.745 0.221
[−0.594, 2.084] [−0.987, 1.424]
Master's degree or higher 0.602 0.275
[−0.729, 1.935] [−0.933, 1.477]
Primary Education 1.053 0.677
[−0.327, 2.447] [−0.590, 1.908]
Secondary Education 0.483 0.146
[−0.838, 1.810] [−1.045, 1.352]
Vocational School 0.504 0.048
[−0.834, 1.837] [−1.154, 1.248]
Ideology
Ideology 0.327 0.319
[0.194, 0.460] [0.189, 0.448]
Minority
Minority: Yes −0.070 0.075
[−0.251, 0.110] [−0.102, 0.250]
Minority: Decline −0.374 −0.182
[−0.743, 0.003] [−0.539, 0.170]
Gender
Female −0.720 −0.768
[−0.857, −0.584] [−0.901, −0.636]
None of the Above −2.734 −1.547
[−5.533, −0.414] [−4.289, 0.687]
Intercept
Num.Obs. 2172 2178
R2 0.102 0.136
R2 Marg. 0.101 0.135
ELPD −2816.7 −3228.4
ELPD s.e. 33.9 24.4
LOOIC 5633.5 6456.7
LOOIC s.e. 67.8 48.9
WAIC 5631.2 6454.4
N 2172 2178
N.Groups 16 16
Groups province province
sd.Province. 0.05 0.064

References

Allen, Michael A., Michael E. Flynn, Michael E. Martinez Machain, and and Andrew Stravers. 2022. Beyond the Wire: US Military Deployments and Host Country Public Opinion. Oxford, New York: Oxford University Press.
Allen, Michael A, Michael E Flynn, Carla Martinez Machain, and Andrew Stravers. 2020. “Outside the Wire: US Military Deployments and Public Opinion in Host States.” American Political Science Review 114(2): 326–41.
Bürkner, Paul-Christian. 2017. “Brms: An R Package for Bayesian Multilevel Models Using Stan.” Journal of Statistical Software 80: 1–28.
———. 2018. “Advanced Bayesian Multilevel Modeling with the R Package Brms.” The R Journal 10(1): 395–411.
Kruschke, John K. 2015. Doing Bayesian Data Analysis. Second. New York, NY: Academic Press.